Apparently, unordered_set::erase and unordered_set::count return something that is not strictly boolean (logically, that is, I'm not talking about the actual type).
The linked page reads for the third version of erase:
size_type erase( const key_type& key );
Removes the elements with the key value key
This has a tone to it that suggests there could be more than just one element with a given key. It doesn't explicitly state this, but it sounds like it a lot.
Now, the point of a set, even an unordered one, is to have each element once.
The standard library acknowledges the existence of the bool type and uses it for boolean values like unordered_set::empty(). So, what's the point of returning size_type in the cases above? Even in spite of hash collisions, the container should distinguish elements with different keys, right? Can I still rely on that?
a.erase(k) size_type Erases all elements with key equivalent to k. Returns the number of elements erased.
b.count(k) size_type Returns the number of elements with key equivalent to k.
It's because of the unordered associative container requirements [23.2.5].